Else et a\ Abundance of Sebaslolobus alasconus in the Gulf of Alaska 



195 



were calculated from the position fixes taken from the sur- 

 face vessel. 



The entire length of each transect was viewed and all 

 thornyhcads in the field of view were counted. Inverte- 

 brates were also counted in seven categories: starfish, seap- 

 ens, sea urchins, anemones, corals, sponges, and sea cu- 

 cumbers. Depth and temperature were recorded once every 

 minute during each transect and averaged over the com- 

 plete transect. Substrate type was estimated by scoring vid- 

 eo quadrats at one-minute intei-vals during the transect. A 

 mylar grid of about twenty-five 50-mm squares was placed 

 on the screen over the freeze-framed image, and substrate 

 within the squares was scored in three categories of soft 

 (mud, sand, and gravel), cobble, and rock-boulder, the lat- 

 ter two of which were then combined into a single catego- 

 ry, hard-bottom. The proportions of each category (soft and 

 hard) for each transect were estimated as the mean from 

 the one-minute quadrats. Abundance (number/100 m-) of 

 shortspine thornyheads and invertebrate categories was es- 

 timated for each transect by dividing the number counted 

 by the area estimate ( W x Ti-ansect length x 100). 



To evaluate variables that may have affected thorny- 

 head densities, we assembled a correlation matrix and 

 a partial correlation matrix of three physical variables 

 (depth, substrate, temperature) and transformed (log(.v-i-l)) 

 biotic variables (shortspine thornyheads and seven inver- 

 tebrate categories). Based on those matrices, we selected 

 depth and substrate for further analyses. 



We used nonparametric procedures (Kruskal-Wallis and 

 Mann-WTiitney) to determine if thornyhead densities var- 

 ied among sampling stations and to evaluate the effects 

 of substrate and depth. Substrate and depth were coded 

 into nominal categories for use as independent variables 

 in those analyses. We used Scheffe's (Zar, 1984) post-hoc 

 test to evaluate between-categorv differences when Krus- 

 kal-Wallis results were significant. 



An ANOVA factorial model was used to explore the 

 joint effects depth and substrate on shortspine thorny- 

 head abundance (transformed by logtx-i-l)). In addition, 

 we used stepwise linear regression with thornyhead abun- 

 dance as the dependent variable to evaluate the relative 

 importance of all possible independent variables. 



0.2 0.3 0.4 0.5 0,6 0.7 0.8 0.9 

 Distance (km) 



50 B 



40 

 c 30 



0) 



o 



I 20 



10- 







100 150 200 250 300 350 400 450 



Depth (m) 



0.4 0.6 O.i 

 Hard substrate 



Figure 2 



Distribution of transects by (A) distance (km). 

 (B) average depth, and (C) the proportion of 

 bottom categorized as hard. 



Results 



Transect lengths ranged from 320 to 800 m and had a 

 mean of 580 m (Fig. 2A). The mean transect depth ranged 

 from 165 to 355 meters, and the highest numbers of tran- 

 sects were at 200-250 m (Fig. 2B). Most transects (85'7f ) 

 had substrate that was completely soft (silt, sand, and 

 gravel) or hard (cobble and rock-boulder) (Fig. 2C). Abun- 

 dance of shortspine thornyheads per transect ranged from 

 to 13.6/100 m'-; and the mean abundance at the 27 sta- 

 tions ranged from to 7.5/100 m- i Table 1). Difference in 

 abundance among stations was very highly significant (P 

 <0.0001, Kruskal-Wallis). 



A correlation matrix of all variables indicated that 

 depth, substrate, and sponge abundance were related to 

 variation in thornyhead abundance; however, the partial 



correlation matrix indicated that among those three vari- 

 ables, substrate type and depth were most strongly re- 

 lated to shortspine thornyhead abundance (Table 2). The 

 high correlation of sponge and thornyhead abundances 

 was apparently spurious because of the relationship be- 

 tween sponge abundance and substrate type (Table 2). 



For further analyses of the relationship between thorny- 

 head abundance, depth, and substrate, we coded depth into 

 three nominal categories, <200 m, 200-300 m, and >300m, 

 chosen to correspond to the depth intervals used in National 

 Marine Fisheries Service (NMFS) triennial trawl surveys 

 (Stark and Clausen. 1995). We also coded substrate into two 

 nominal categories, soft bottom (>90'7( sand, mud, and grav- 

 el) and hard bottom (>40% cobble and rock-boulder). Abun- 

 dance increased with depth (Table 3), and differences that 

 were highly significant occurred among three depth catego- 

 ries (P<0.001, Kruskal-Wallis). Significant differences exist- 



